Create add_dataset.md

Author: lhoestq

.github/PULL_REQUEST_TEMPLATE/add_dataset.md

@@ -0,0 +1,16 @@
+- **Name:** *name of the dataset*
+- **Description:** *short description of the dataset (or link to social media or blog post)*
+- **Paper:** *link to the dataset paper if available*
+- **Data:** *link to the Github repository or current dataset location*
+- **Motivation:** *what are some good reasons to have this dataset*
+
+### Checkbox
+
+- [ ] Create the dataset script `/datasets/my_dataset/my_dataset.py` using the template
+- [ ] Fill the `_DESCRIPTION` and `_CITATION` variables
+- [ ] Implement `_infos()`, `_split_generators()` and `_generate_examples()`
+- [ ] Make sure that the `BUILDER_CONFIGS` class attribute is filled with the different configurations of the dataset and that the `BUILDER_CONFIG_CLASS` is specified if there is a custom config class.
+- [ ] Generate the metadata file `dataset_infos.json` for all configurations
+- [ ] Generate the dummy data `dummy_data.zip` files to have the dataset script tested and that they don't weigh too much (<50KB)
+- [ ] Add the dataset card `README.md` using the template : fill the tags and the various paragraphs
+- [ ] Both tests for the real data and the dummy data pass.